Research On Cloud Computing Elastic Model And Cloud Platform Optimization Method

As a new computing method of on-demand distribution,cloud computing's important feature is dynamic scalability,which enables cloud computing platforms to dynamically expand or shrink cloud resources according to user needs changes.Dynamic expansion according to the user's computing needs is called elasticity,which is the key feature of cloud computing and the key advantage of cloud computing compared to other traditional computing methods.At present,there are two key problems in the cloud computing platform.One is the lack of a model that can represent and quantify the elasticity of the cloud computing platform,,and the other is the lack of effective methods to predict and optimize the performance and cost of the cloud platform.In order to solve the above two problems,this thesis carries out the following research work.(1)Aiming at the lack of elastic model in the cloud computing platform that can be modeled and quantitative analysis,this thesis proposes a new elastic computing model.The model is based on the existing elastic computing model,and takes into account hot virtual machine start-up rate,cold virtual machine start-up rate,hot virtual machine shut-down rate,cold virtual machine shut-down rate,virtual machine service rate,and task reach rate in the actual cloud platform operation.It views the cloud platform as a queuing system to study elastic computing,and uses the queuing model to analyze the cloud platform,and accurately calculates the elasticity of the cloud computing platform through the parameter-based numerical analysis method.(2)Aiming at the lack of effective methods in cloud computing platform to predict and optimize the performance and cost of cloud platform,this thesis applies the above elastic model and queuing theory to predict and optimize many important attributes of elastic computing systems,including system configuration status,expected number of tasks,expected queue length,expected task response time,number of expected busy virtual machines,number of expected idle virtual machines,and multiple attributes expected to shut-down virtual machines.(3)Finally,this thesis proposes a performance and cost optimization method based on multi-objective particle swarm optimization to solve Pareto optimal solution and the corresponding optimal system-level parameters,and uses the optimal stopping strategy to obtain the optimal number of hot virtual machines,and combines non-dominant vector set to obtain the optimal values of the parameters that affect the performance and cost of the cloud platform.Then it analyzes the influence of the changes of the basic parameters and the changes of the combination parameters of the system on the system performance and cost.The elastic model proposed in this thesis considers more influencing factors,which are crucial in the actual cloud computing platform.The research has two meanings.On the one hand,cloud service providers can use the research results of this thesis to predict the performance and cost of cloud computing platforms.On the other hand,cloud service providers can optimize the elastic expansion scheme of cloud computing platforms to provide users with the highest price/performance ratio.